High-frequency MHz-order vibration enables cell membrane remodeling and lipid microdomain manipulation

Abstract

We elucidate the mechanism underpinning a recently discovered phenomenon in which cells respond to MHz-order mechanostimuli. Deformations induced along the plasma membrane under these external mechanical cues are observed to decrease the membrane tension, which, in turn, drives transient and reversible remodeling of its lipid structure. In particular, the increase and consequent coalescence of ordered lipid microdomains leads to closer proximity to mechanosensitive ion channels—Piezo1, in particular—that, due to crowding, results in their activation to mobilize influx of calcium (Ca²⁺) ions into the cell. It is the modulation of this second messenger that is responsible for the downstream signaling and cell fates that ensue. In addition, we show that such spatiotemporal control over the membrane microdomains in cells—without necessitating biochemical factors—facilitates aggregation and association of intrinsically disordered tau proteins in neuroblastoma cells, and their transformation to pathological conditions implicated in neurodegenerative diseases, thereby paving the way for the development of therapeutic intervention strategies.

Significance

Beyond uncovering the fundamental mechanism of a recent and surprising phenomenon, which has been elusive to date, the ability to structurally manipulate the cell membrane constituents through high-frequency vibration, which we have discovered to be central to the mechanism, is shown to have a direct influence on the cell’s eventual fate. By demonstrating that this, as an example, can alter the state of a certain protein that has been implicated in neurodegeneration (tau) in a similar way to how it changes when transitioning to a disease state, this platform can potentially be used to further our understanding of the molecular and cell dynamics associated with disease progression, and hence inform the development of new strategies to treat the disease.

Introduction

The plasma membrane—formed from the self-assembly of amphiphilic lipids, transmembrane and membrane-anchored protein clusters, and carbohydrates—comprises the physical boundary that separates and regulates the interactions between the interior of a cell from its external environment. Due to lipid-lipid and lipid-protein interactions, liquid-liquid phase separation results in the heterogeneity of these lipid and protein assemblies, which manifest as distinct microscopic lateral clusters ($>$300 nm). These clusters can be separated into relatively ordered liquid domains ($L_o$) known as lipid rafts that contain saturated lipids (phospholipids, glycolipids, and, predominantly, sphingolipids) and sterols (mainly cholesterol) (1,2,3,4,5), and disordered clusters ($L_d$) containing unsaturated lipids, with transmembrane proteins partitioned between both regions (6). It is the dynamic spatial reorganization of these regions in response to alterations in the membrane tension as a consequence of deformations imposed on the plasma membrane that enable its role as a primary mechanosensor (7,8,9), which, in turn, plays a crucial role in signal transduction activation that is the determinant of downstream cellular fate.

Despite their biological significance, the ability to directly manipulate the lipid microdomains on the plasma membrane—which could not only provide the elusive tool to study lipid rafts but also enable a potent means for engineering cells—remains challenging without modifying lipid composition, osmolarity, or the underlying substrate, or inducing vesicle fusion, all of which fall short of representing key biological processes that recapitulate typical in vivo conditions (10). While local heating with a laser allows the possibility for inducing reversible microdomain transition between the $L_o$ and $L_d$ phases, this has only been demonstrated in model lipid bilayers (i.e., vesicles) (11).

Recently, we had observed that cells exhibit specific responses when subject to high-frequency ($\mathcal{O}$(10 MHz)) mechanostimulation (12), which, in itself, is surprising [most studies, in fact, have, on the contrary, attempted to demonstrate that ultrasound does not cause any noticeable effects on cells (13)] given that such frequencies are far beyond the typical perception range (several Hz, commensurate with that associated with physiological motion) expected of cells [We note that several phenomena can accompany the sound wave excitation that could have an effect on the cell. With the exception of acoustic radiation pressure, which occurs over timescales commensurate with the inverse of the excitation frequency (in the present case, ${10}^{-7}$ s), other effects such as acoustic streaming or bubble resonance possess characteristic frequencies considerably distinct to the excitation frequency. Their characteristic timescales nevertheless tend to still be much smaller than the macroscopic timescale over which the cell typically responds. As we discuss subsequently, however, it should be noted that bubble resonance is irrelevant as cavitation is essentially nonexistent in our system; moreover, we operate at sufficiently low powers to avoid acoustic streaming and hence effects of fluid shear on the cell]. Central to all of these responses and the downstream cell fates they induce (e.g., cell adhesion, migration and proliferation (14,15), vesicle trafficking and exosome biogenesis (16), immune cell activation (17), stem cell differentiation (18,19), endothelial barrier modulation (20), and neuromodulation (21,22)) are the transcriptomic changes brought about by second messenger signaling that are instigated by the ability for the external high-frequency mechanical cues to modulate Ca²⁺ levels within the cell through the activation of various mechanosensory elements along the plasma membrane, such as mechanosensitive ion channels and membrane-associated proteins (12).

In this work, we report our attempt to uncover the fundamental mechanisms responsible for this signal transduction process. In doing so, we shed light on the processes by which the high-frequency mechanostimulation, applied for just several minutes, drives transient deformation of the cell and reorganization of its cytoskeletal structure through a dynamic combination of compression, tension, and shear. In particular, we show that the resultant change in the cell’s membrane tension as a consequence leads to spatiotemporal remodeling of the lipid structure and hence changes in its microdomain distribution.

Parenthetically, we note that the phenomena we report here due to 10 MHz-order mechanostimulation, in the form of nanometer amplitude surface acoustic waves (typically $>$20 MHz) and hybrid surface and bulk acoustic waves, i.e., surface reflected bulk waves (SRBWs; typically $<$20 MHz) (23), are considerably distinct from that reported with 1 kHz-order vibration (24,25,26), and, more generally, any mechanostimuli conducted at frequencies below 1 MHz using focused ultrasound (27). Given that cavitation, which is notably absent at 10 MHz-order frequencies under the intensities we typically employ (28), becomes increasingly prevalent at lower frequencies, mechanostimulation at these lower frequencies has primarily been focused on cavitation-induced poration effects (i.e., sonoporation), and hence it is likely that the long continuous exposure durations typically employed, and the accompanying poration that ensues, results in uncontrolled disruption of the membrane and its lipid microdomains to facilitate other downstream fates (27,29,30,31). As such, the effects we observe in the present work due to the high-frequency mechanostimulation, particularly its ability to directly manipulate the lipid microdomains on the plasma membrane, do not appear to have been reported at lower frequencies. In addition, we note that the observed phenomena in this work cannot be attributed to flow-induced shear since the intensities we typically employ are below those required to generate appreciable acoustic streaming (i.e., the flow arising as a consequence of viscous dissipation as the sound waves attenuate in the liquid medium).

Motivated by such potential to activate ordered membrane-associated proteins (such as Piezo1) through spatiotemporal lipid microdomain manipulation, we show, in the final section, the intriguing possibility for transforming intrinsically disordered proteins with the high-frequency mechanostimulation. In particular, we demonstrate the transformation of tau proteins (tubulin-associated units), whose function and structure depend on their interactive partners (membranes, proteins, or RNA), and whose association with the plasma membrane and aggregation has been implicated, albeit contentiously, in cell senescence in the brain, and hence the pathology of neurodegeneration (e.g., Alzheimer’s disease and frontotemporal dementia (22,32,33)).

Materials and methods

Materials

Human mesenchymal stem cells (hMSCs) were acquired from Lonza (Mount Waverley, VIC, Australia), whereas the SH-SY5Y neuroblastoma cell line was obtained through Dr Shwathy Ramesan at The Florey Institute of Neuroscience and Mental Health (Parkville, VIC, Australia). Silicone oil, Triton X-100, methyl-β-cyclodextrin, dimethylsulfoxide (DMSO), Opti-MEM phosphate-buffered saline (PBS), Gibco penicillin-streptomycin, trypsin-ethylenediaminetetraacetic acid (EDTA), formaldehyde, bovine serum albumin (BSA), fetal bovine serum (FBS), Dulbecco’s modified Eagle’s medium (DMEM), Eagle’s minimum essential medium, F12 medium, Fura-2 acetoxymethyl ester (Fura-2AM), 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid tetrakis(acetoxymethyl ester) (BAPTA-AM), trypan blue solution, cholera toxin subunit B (recombinant; CtxB), ActinRed 555 ReadyProbes, NucBlue Live ReadyProbes, Alexa Fluor 647 conjugate, anti-Piezo1 antibody, T25 cell culture flasks, and Nunc Lab-Tek II Chambered Coverglass were sourced from Thermo Fischer Scientific (Scoresby, VIC, Australia). FluoroDish cell culture plates were procured from World Precision Instruments (Hertfordshire, England). Anti-tau rabbit monoclonal antibody, and anti-rabbit and anti-mouse IgG (H+L) F(abʹ)2 fragment (Alexa Fluor 488 conjugate), on the other hand, were obtained from Cell Signaling Technology (Danvers, MA), whereas C-Laurdan and Flipper-TR (CYT-CY-SC020) were obtained from R&D Systems (Minneapolis, MN) and Cytoskeleton (Denver, CO), respectively.

Device fabrication

The SRBW devices shown in Fig. 1 a consisted of 40 alternating pairs of 11-mm-wide and 66-nm-thick straight interdigitated aluminum transducer (IDT) electrodes arranged in a basic full-width interleaved configuration on 500-$\mu$m-thick ${127.86}^{\circ }$ Y-X rotated lithium niobate (LiNbO₃) single-crystal piezoelectric substrates (Roditi, London, UK). Sputter deposition and standard ultraviolet photolithography were used to pattern the IDTs on top of a 33-nm-thick chromium adhesion layer. The resonant frequency of the device $f=c/\lambda$, where c is the SRBW phase speed in LiNbO₃, was set at 10 MHz by prescribing the width and gap of the IDT fingers (λ/4). To generate the SRBW, a signal generator (SML01; Rhode & Schwarz, North Ryde, NSW, Australia) and amplifier (10W1000C; Amplifier Research, Souderton, PA) were used to apply an alternating electrical signal to the IDT at the resonant frequency. The acoustic wave energy from the device was transmitted to the cells adherent to the glass-bottom chamber slide through a thin layer of silicone oil (34) with a viscosity of 45–55 cP and density 0.963 g/mL at 25°C.

Figure 1.

Experimental setup and plasma membrane tension changes in response to the SRBW forcing. (a) Top and side view schematics showing the experimental setup in which the SRBW (not to scale), generated along a piezoelectric lithium niobate (LiNbO₃) substrate by applying an AC electric signal at the device’s resonant frequency (10 MHz) to an interdigitated transducer electrode (IDT) photolithographically patterned on the substrate, is coupled through a thin layer of silicone oil into a glass-bottom cell culture plate containing the adherent cells. (b) Schematic representation of the changes in the plasma membrane tension due to the SRBW irradiation, measured using the probe Flipper-TR, which, when tagged to the membrane lipids, allows quantification of the pressure along its axis: “sonochallenged” cells (i.e., the state of the cells immediately after cessation of the SRBW irradiation [0 min postexposure incubation]) exhibited a reduced fluorescence lifetime similar to cells under hyperosmotic conditions, while “sonotransformed” cells (i.e., the state of the cells after their recovery [20–60 min postexposure incubation]) regained higher fluorescence intensity lifetime values close to that of untreated (control) cells owing to relaxation of the membrane. (c) Representative images (color coded for the intensity-weighted averaged fluorescence lifetime whose corresponding values are presented in Fig. S1) of Flipper-TR-probed hMSCs exposed to 8 min of SRBW irradiation at input powers of 0, 1.5, and 2.5 W and fixed at different postexposure incubation durations; control untreated cells (0 W) were fixed at 0, 20, and 60 min, while cells irradiated at 1.5 and 2.5 W were fixed at 0, 10, and 20 min and 0, 20, and 60 min, respectively, as determined through a series of optimization studies. Scale bars, 40 $\mu$m. Comparison of (d) the longest component associated with the best fit to the fluorescence lifetime curve ${\tau }_1$, (e) the cell area, (f) the average number of protrusions per cell, and (g) the relative filipodia length in untreated and SRBW-irradiated cells; the data are represented in terms of its mean $\pm$ standard error over triplicate runs, and the asterisks ^∗∗∗ and ^∗∗∗∗ indicate statistically significant differences with $p<0.001$ and $p<0.0001$, respectively.

Cell culture and mechanostimulation

The hMSCs were grown in DMEM with 1% penicillin-streptomycin and 10% FBS in a humidified incubator at 37°C and 5% CO₂ until they covered 80–90% of a standard T25 flask. They were then detached using 0.05% trypsin-EDTA and transferred to 8-well plates or FluoroDish cell culture plates at a seeding density of 3000 cells per well before being incubated for 28 h in DMEM to ensure proper adhesion. SH-SY5Y neuroblastoma cells, on the other hand, were expanded in a 1:1 mixture of Eagle’s minimum essential medium and F12 medium with 10% FBS and 1% penicillin-streptomycin.

The cells in the well plate were subsequently exposed to 10 MHz SRBW irradiation at different input powers and exposure times. Unless otherwise specified, the exposure time was 10 min for input powers below 2 W, and 5 min for input powers of 2 and 2.5 W. The cells were incubated immediately for different periods after cessation of the SRBW irradiation before being processed for further analysis. Control samples consisted of cells seeded in DMEM with 1% penicillin-streptomycin and 10% FBS at the same density and incubated for the same time period, but without any vibrational excitation.

For the inhibitor studies, the SHSY-5Y cells were seeded at 3000 cells per well and incubated with BAPTA-AM (10 $\mu$M, for 25 min) or methyl-β-cyclodextrin (300 $\mu$M, for 60 min) before SRBW irradiation and washed with medium devoid of the inhibitors. The irradiated cells were then fixed after 5 min of postexposure incubation. The control comprised cells unexposed to SRBW stimulation but incubated with medium containing the same concentration in DMSO.

Membrane tension

The process for characterizing the membrane tension of the cells is as follows. The cells were first incubated with 3 $\mu$M Flipper-TR in DMSO for 15 min at 37°C, and subsequently washed three times in medium devoid of the stain. The cells were then exposed to the SRBW mechanostimulation at different input powers (1.5 and 2.5 W), after which they were fixed at various postexposure incubation times. Fluorescence lifetime imaging (FLIM) was subsequently carried out with an inverted confocal laser scanning microscope (FV3000; Olympus, Tokyo, Japan) equipped with a time-correlated single-photon counting module (PicoQuant, Berlin, Germany) to measure the lifetime of the probe. The FLIM excitation source comprised a pulsed 485 nm laser (LDH-D-C-485; PicoQuant) operating at 20 MHz, and the emission signal was collected through a 600/50 nm bandpass filter using a gated hybrid photomultiplier detector (PMA Hybrid 40; PicoQuant) and a time-correlated single-photon counting and multichannel scaling board (MultiHarp 150; PicoQuant). The intensity-weighted averaged lifetimes of the probe were then determined from the acquired FLIM images using the supplied software (SymPhoTime 64; PicoQuant). The area of the cells and the protrusion-like structures, on the other hand, were measured using ImageJ (National Institutes of Health, Bethesda, MD), the latter using the plugin FiloQuant; examples of how the cell boundaries and protrusions were defined for the analysis using the plugin are shown in Fig. S2.

Membrane polarization

Membrane polarization was assessed by first treating the cells with 3 $\mu$M C-Laurdan in DMSO, and washing them three times in medium devoid of the stain. The cells were then exposed to the SRBW mechanostimulation at various input powers (0.5, 1, 1.5, 2, and 2.5 W). All cells, except those exposed to the SRBW at 2.5 W, were fixed at the end of the 10-min exposure duration. The cells that were exposed to the SRBW at 2.5 W (5 min) were fixed at various postexposure periods (0, 20, and 60 min). Two-photon fluorescence images were captured using a confocal microscope (FV3000; Olympus) equipped with a $100\times$ oil immersion objective (UPLAPO100XOHR, NA 1.5; Olympus). A femtosecond Ti:Sapphire laser (InSight X3; Spectra-Physics, Milpitas, CA) operating at 780 nm was used for optical excitation. The optical power at the sample stage was adjusted to 4 $\mu$W, and intensity images were recorded across two channels with emission in the range of 419–457 and 575–525 nm. The generalized polarization (GP) values were determined using the ImageJ macro (National Institutes of Health) specified in (35), along with a sensitivity correction factor of 0.3, which was established by measuring the relative sensitivities of the two channels. GP distributions were derived from normalized frequency histograms and fitted to one or two Gaussian functions with a nonlinear fitting algorithm (Origin 7.0; OriginLab, Northampton, MA). Only fits for which the chi-squared test yielded values of $p>0.95$ were considered acceptable. The area under the peaks and the maximum peak fit values were, respectively, used to define the coverage (relative percentage) of the $L_o$ and $L_d$ microdomains, and the high and low GP values for each individual experiment, which was independently replicated at least three times. Images displayed in the figures are representative of the wider data set that was used for quantification.

Immunofluorescence staining

The cells requiring fixation were washed three times in PBS and incubated in 4% formaldehyde for 20 min at room temperature, followed by three further wash cycles in PBS. The cells were then blocked by incubating them in 5% BSA in PBS for 60 min, followed by further incubation with CtxB Alexa Fluor 647 conjugate overnight at 4°C. The cells were subsequently washed three times in PBS, permeabilized with 0.1% Triton X-100 for 5 min, washed a further three times in PBS, and then blocked again with 5% BSA in PBS for 60 min, after which they were incubated with anti-Piezo1 (1:500) or anti-tau (1:500) antibodies overnight at 4°C. After three washes in PBS, the cells were next incubated in the secondary antibody (anti-rabbit and anti-mouse IgG (H+L) F(abʹ)2 fragment [Alexa Fluor 488 conjugate]; 1:1000) for 1 h in the dark at room temperature and subsequently washed three more times in PBS. Nuclei were counterstained using NucBlue Live ReadyProbes, while actin filaments were stained with ActinRed 555 ReadyProbes before imaging under a confocal microscope (A1 HD25; Nikon Instruments, Melville, NY). Lipid raft (ganglioside [GM1]) and Piezo1 colocalization distances were measured using the ImageJ distance analysis plugin DiAna (National Institutes of Health), whereas the distribution studies in Fig. 3 a were analyzed using the plot profile in ImageJ (National Institutes of Health). Fig. S3 shows examples of how the cellular boundaries and protrusion-like features were defined for the analysis.

Figure 3.

Lipid raft and Piezo1 reorganization under the SRBW forcing. (a) Representative images depicting the distribution of lipid raft residing gangliosides (GM1) and Piezo1 channels in untreated (control) and SRBW-irradiated (2.5 W) cells at different postexposure incubation periods. Cell nuclei were stained using NucBlue Live ReadyProbes (blue), GM1 with Alexa Fluor 647 conjugate (purple), and Piezo1 with anti-Piezo1 antibody and Alexa Fluor 488 secondary antibody (green). The third column comprises a merger of the channels, whereas the last two columns comprise enlarged views of the region indicated in the merged channel. Scale bars, 50 $\mu$m. The corresponding line intensities of the GM1 (purple) and Piezo1 (green) fluorescent signals with respect to that for the nucleus (blue) is shown on the right. (b) Frequency distribution ($n=150$) of the distance between GM1 and Piezo1 in untreated and SRBW-irradiated (2.5 W) cells at different postexposure incubation times. (c) Intracellular ${\text{Ca}}^{2+}$ ($n=6$) as a function of the SRBW power. (d) Images showing actin reorganization in untreated and SRBW-irradiated cells (2.5 W for 5 min) at different postexposure incubation times; the actin structures were stained with ActinRed 555 ReadyProbes (red) and cell nuclei were stained with NucBlue Live ReadyProbes (blue). Scale bars, 50 $\mu$m. (e) Schematic illustration showing redistribution of the lipid raft and Piezo1 channels under the SRBW forcing: untreated cells or those initially at ground state before the SRBW irradiation comprised a heterogeneous microdomain composition, whereas sonochallenged cells immediately after SRBW irradiation displayed increasing homogeneity in $L_o$ microdomains, whose coalescence leads to their clustering with membrane-associated proteins such as Piezo1, as evident by their increased colocalization, which, in turn, results in Piezo1 activation and an increase in intracellular ${\text{Ca}}^{2+}$ into the cell. The data are represented in terms of its mean $\pm$ standard error over triplicate runs, and the asterisks ^∗∗∗∗ indicate statistically significant differences with $p<0.0001$.

Intracellular ${\text{Ca}}^{2+}$

The fluorescent calcium indicator Fura-2AM was used to measure free cytosolic ${\text{Ca}}^{2+}$ levels. In brief, the cells were placed in 5 $\mu$mol/L Fura-2AM in Opti-MEM reduced serum medium with 2% (vol/vol) heat-inactivated FBS and incubated at 37°C in the absence of light. After 1 h of incubation, the medium with extracellular Fura 2AM was replaced with fresh medium followed by further incubation for an additional 20 min before irradiating the sample with the SRBW. Using a spectrophotometric plate reader (CLARIOstar, BMG LabTech, Mornington, VIC, Australia), we subsequently measured the fluorescence emission intensity at 510 nm in individual wells at excitation wavelengths of 340 and 380 nm. The ratio of Fura-2AM fluorescence emission in response to 340 and 380 nm excitation (340/380) can then be calculated and expressed as the fold change relative to that measured for the respective control (unexcited) cells.

Statistics

Quantitative data are presented as the mean $\pm$ standard deviation from a minimum of three independent experiments. Statistical analysis involved one-way analysis of variance; a threshold of $p<0.05$ signified statistical significance.

Results and discussion

As model cell lines, hMSCs from bone marrow and SH-SY5Y neuroblastoma cells cultured in glass-bottom cell culture plates were exposed to 1.5 and 2.5 W SRBW, as shown in the experimental setup illustrated in Fig. 1 a; these powers were chosen based on optimization studies that reveal the most significant mechanoresponse for these particular cells. We then probed changes in their membrane tension and lipid structure at different postincubation periods using various assays, as described in the Materials and methods.

Deformation-driven changes to the plasma membrane tension

Given the considerably longer SRBW wavelength at 10 MHz ($\lambda =398\mu$m) but much smaller amplitude $\mathcal{O}$(10 nm) compared with typical cellular dimensions, and as the SRBW vibrational forcing timescale ($1/f\sim \mathcal{O}({10}^{-7}$ s)) far exceeds characteristic cellular timescales ($\mathcal{O}$(1–100 s) (36)), it would be reasonable to expect that cells are unlikely to deform under such high-frequency mechanostimuli, at least under leading order modes and over timescales commensurate with the inverse of the frequency. Yet, considerable deformation is seen, as evident from the changes to the cells’ membrane tension observed when they are exposed to the SRBW. Figs. 1 c, d and S1, for instance, show a marked decrease in membrane tension immediately upon exposure to the SRBW at input powers of 1.5 and 2.5 W, compared with that for unexposed (control) cells, as captured through the fluorescence lifetime of the membrane tension probe Flipper-TR—a planarizable push-pull probe which, once inserted into the membrane lipid, undergoes conformable (planar versus twist) changes that influence the lifetime of its excited state (Fig. 1 b)—bound to the cells. Specifically, the tension in the plasma membrane correlates linearly with the longest component of the fluorescence lifetime ${\tau }_1$, which, in turn, is dependent on the intramolecular charge transfer that occurs during changes in the twist angle and polarization between the two twisted dithienothiophenes of the mechanophore (37,38,39). Owing to the short incubation periods in our experiments ($<$2 h), the ${\tau }_1$ values measured correspond solely to effects arising from the plasma membrane tension and not those due to changes in the intracellular structures. Importantly, we observe the mechanostimulated cells to eventually relax back to their original state; the larger the applied SRBW power, the longer the period over which this occurs, with the cells exposed to 1.5 W recovering within 20 min and those exposed to 2.5 W recovering within 60 min. This transient effect is consistent with observations in our previous studies, be it for intracellular delivery wherein such transient effects, together with other factors, provided strong evidence that the SRBW did not result in physical pore formation, unlike in sonoporation (40,41); or with endothelial cells, where barrier function was observed to be uniquely recovered after its initial permeabilization following the sonochallenge—such recovery not being observed with chemical insults to cells (20), for example.

The decrease in membrane tension as a consequence of the SRBW mechanostimulation is akin to that observed in cells under hyperosmotic conditions (38,42,43); we note that the levels to which the membrane tension decreases due to the SRBW forcing, as captured by the reduction in ${\tau }_1$ by approximately 0.5–1 ns, is similar to that reported for these cells (38), or those mechanostimulated by static compression (38,44). In both these cases (i.e., cells exposed to hyperosmotic conditions or static compression), distinct morphological traits, such as contraction of the cell, a two-dimensional drum-like appearance on its apical side, and protrusions that resemble filopodial structures, were reported, all of which have been observed with cells exposed to the SRBW mechanostimulation (40). Given that these morphological traits appear to form immediately after the cessation of the stimulation, it is likely that they arise due to the retraction of the membrane toward the center of the cell or nucleus, as indicated by the transient yet quick reduction in cell size (Fig. 1 e), the increase in the number and length of protrusions (Fig. 1 f and g), and the distribution of the membrane stain distinctly around the nucleus as a response to its contraction under the mechanostimulation, which can be seen to decrease toward the cell periphery (Fig. 1 c). In any case, all of the characteristics observed can be seen to be more prominent at the higher SRBW intensity (i.e., 2.5 W), but in a manner akin to the recovery of the membrane tension to levels associated with its original state at rest, diminishes with postexposure incubation time (Fig. 1 c).

Effect of changes in membrane tension on membrane fluidity

In contrast to cellular timescales, characteristic membrane lipid relaxation times ($\mathcal{O}$(1 ps) (45)) are much shorter than the timescale associated with the SRBW forcing. As such, and given that the deformation-induced decrease in membrane tension is likely to alter the membrane fluidity and hence the homogeneity of its domain composition, it is not unreasonable to then expect modifications to the lateral organization of the membrane lipid structure to accompany the SRBW mechanostimulation process. Indeed, we observe in Fig. 2 that the SRBW mechanostimulated cells, stained with C-Laurdan—a commonly used probe for determining membrane lipid order—display a notable increase in their GP values (lower GP values are typically associated with the dominance of $L_d$ domains wherein the plasma membrane has a closer resemblance to a fluid-like state, whereas higher values are associated with the dominance of $L_o$ domains in which the plasma membrane resembles a stiffer gel-like state (46,47,48,49); Fig. 2 a), which is indicative of an increase in the relative proportion between the ordered $L_o$ and disordered $L_d$ phases (46) (the $L_o/L_d$ ratio being characteristic of the fluidity of the membrane and can vary between cell types and passages). More specifically, we note from Fig. 2 b a shift toward higher GP values as the plasma membrane transitions toward a more homogeneous gel-like state with increasing SRBW power, compared with the unexposed control, which exhibited a more heterogeneous composition comprising both $L_o$ and $L_d$ domains. Consistent with our earlier observations for the membrane tension, however, we observe in Fig. 2 c and e the peak distribution to shift back to values corresponding with the cells’ ground state with time: the heterogeneity in the domain composition with both low and high GP values can be seen to reappear within 60 min postincubation exposure for the SRBW mechanostimulated cells at 2.5 W.

Figure 2.

Change in membrane fluidity due to the SRBW irradiation. (a) Schematic representation showing the effect of the SRBW on membrane fluidity: cells at ground state (and untreated cells) in which both $L_o$ and $L_d$ domains coexist (left), giving rise to both low and high generalized polarization (GP) components; increased membrane fluidity in sonochallenged cells result in a transition to a more gel-like phase and hence higher GP values (center); at longer postexposure durations, sonotransformed cells return toward their ground state through increased exocytosis (right). (b and c) Images (inset) and distribution of their GP values ($-1$ to $+1$; as represented by the color bar) for control and SRBW-irradiated C-Laurdan-labeled hMSCs, (b) fixed at the end of the exposure period, as a function of the SRBW power, and (c) fixed at different postexposure incubation periods for an SRBW power of 2.5 W. The histograms for the GP values were fitted with Gaussian distributions, with the solid green curves representing the average between low (dashed blue curves) and high (dashed red curves) distributions. Scale bars, 30 $\mu$m. (d and e) Average GP values in the cases shown in (b) and (c), respectively. The data are represented in terms of its mean $\pm$ standard error over triplicate runs, and the asterisks ^∗∗, ^∗∗∗ and ^∗∗∗∗indicate statistically significant differences with $p<0.01$, $p<0.001$ and $p<0.0001$, respectively.

The increase in GP values, and hence membrane order and packing, with the SRBW mechanostimulation is, however, contrary to that observed for cells subject to hyperosmotic treatment, where decreases in membrane tension led to a decrease in $L_o$ domains (38). Similarly, our observations here are also contrary to that reported with increasing temperatures (46), therefore suggesting that the effect is unlikely a consequence of heating (which, in any case, is minimal given the low SRBW powers and short exposure duration of just a few minutes), consistent with our previous studies that verify that the SRBW mechanostimulation did not lead to the production of heat shock proteins or other heat-related cell distress (40). We should note that, while our result showing a decrease in membrane fluidity to accompany a decrease in membrane tension may appear to be counterintuitive, they are nevertheless not without precedent, with the literature having reported similar observations (38,50,51,52,53,54,55,56), and suggesting that the membrane lipid reorganization dynamics may be considerably more complex than that afforded by a simplistic prima facie conventional mechanics standpoint (7,57).

The dominance in $L_o$ phases observed nevertheless resembles that observed in giant unilamellar vesicles with compositions within the $L_o/L_d$ fluid coexistence region (53,57,58). In particular, we note that just several minutes of SRBW mechanostimulation induces a transition toward a more gel-like phase that is also accompanied by a change in the $L_o$ coverage, primarily due to fusion or coalescence of the $L_o$ domains (59,60), as seen in Fig. 2 b; this is in contrast to the slow coalescence observed for mechanostimulation at 1 MHz or below, wherein over 45 min of continuous exposure was required, and in which no appreciable morphological changes were observed (27). More importantly, we observe from Fig. 2 b and d, control over the membrane fluidity with the SRBW power, with powers below 1 W leading to a decrease in the $L_o$ domains and hence a transition in the membrane toward a liquid-like state, whereas higher powers beyond 1 W yielding an increase in the $L_o$ domain coverage and a transition in the membrane toward a gel-like state; similar manipulation of the microdomains can be achieved through the postexposure incubation duration, as shown in Fig. 2 c and e. While mechanical cues such as shear, compression, and tension have been reported to facilitate shifts in membrane packing and order, such rapid, direct control over microdomain organization has only been shown to date in giant unilamellar vesicles (11), and not in more complex lipid systems such as cells.

In any case, given the interconnectivity of the plasma membrane with the cytoplasm through the membrane skeleton (61), the SRBW-driven alteration of the lipid microdomains is then anticipated to drive reorganization of the underlying cytoskeletal network. In particular, alterations in the $L_o$ domains or cholesterol content in the plasma membrane manifest in cytoskeletal docking, and thus changes to cell stiffness and morphology. Such changes have previously been observed with SRBW mechanostimulation of both MSCs and endothelial cells, wherein rearrangement of the cytoskeletal structure, namely actin stress fiber formation and remodeling, have been reported (18,20). In particular, the mechanoresponse to the SRBW forcing follows the aforementioned challenge and recovery phases: upon stimulation (i.e., the initial sonochallenge phase), cells were observed to become smaller, stiffer, and rounder, followed by a latent response (within 20–60 min; i.e., the subsequent sonotransformed phase) in which the cells relax to their ground state. The role of actin in facilitating membrane lipid compartmentalization is prominent in both cases. In the former sonochallenge phase, the increase in the actin-associated $L_o$ domains not only leads to more rigid cells; transient coalescence of these (i.e., the $L_o$) domains leads to fusion of $L_d$ domains that lack actin association (62,63,64,65), inducing the membrane to invaginate and to facilitate recruitment of endosomal sorting complexes required for transport (ESCRT), which is responsible for exosome biogenesis (i.e., the formation of intraluminal vesicles and multivesicular bodies) (66,67,68,69), which we had previously observed with SRBW mechanostimulation (16). Furthermore, ESCRT signaling is also responsible for triggering the relaxation of the microdomains toward their ground state in the latter sonotransformation phase. The multivesicular body docking and fusion with the membrane that occurs as a consequence, then results in the reported enhancement in exocytosis (16) that, in turn, relieves the membrane tension; the 30–60 min window over which enhanced exosome secretion is reported (16) strongly corroborating with increases in the extracellular vesicle structures observed in the SRBW-excited cells (Fig. S3), as well as the time over which the plasma membrane returns to its ground state.

Lipid raft association and Piezo channel activation

In the preceding sections, we have shown the SRBW to drive deformations in the plasma membrane leading to its decrease in tension, and, in that process, to facilitate an increase in membrane order and polarization. To complete the picture of how cells respond to the SRBW (and, for that matter, the surface acoustic wave) mechanostimulation, we now show how such reorganization in the microdomains can lead to the myriad of downstream cell fates observed (e.g., cell adhesion, migration and proliferation, vesicle trafficking and exosome biogenesis, immune cell activation, stem cell differentiation, and endothelial barrier modulation) in which the central role of second messenger ${\text{Ca}}^{2+}$ signaling has been implicated. Fig. 3 a and b, in particular, shows the lateral distance along the plasma membrane between the glycosphingolipids (in particular, monosialotetrahexosylgangliosides [GM1] stained with CtxB as a lipid raft marker) and the Piezo channels (in particular, Piezo1), which are mechanically activated ion channels known to facilitate Ca²⁺ transport and to hence initiate its signaling cascades. More specifically, we observe the tendency toward colocalization of gangliosides and Piezo1 after SRBW mechanostimulation, as illustrated in Fig. 3 e and as evident by the increasing proximity between the lipid rafts (i.e., the $L_o$ microdomains) and Piezo1 channels (Fig. 3 b), which were originally distributed throughout the cells. In particular, both can be seen in the figure to move toward and concentrate within the nuclear region.

As observed previously, however, this effect is also transient, with both the lipid rafts and Piezo1 channels returning toward the cell periphery within 60 min, and hence divaricating back toward the average equilibrium separation distances seen for cells in their initial ground state as well as that for the untreated (control) cells (Fig. 3 b). We note that this recovery timescale corroborates previous observations of the role of another second messenger, cyclic adenosine monophosphate (cAMP), that has also been identified as an instigator of key signaling pathways associated with the SRBW mechanotransduction dynamics, wherein elevated intracellular cAMP levels were recorded within 30 min postincubation after SRBW excitation (20).

Whereas increases in membrane tension leading to activation of mechanosensitive ion channels have been reported (see, e.g., (70,71)), there has not been specific work to our best knowledge directly linking decreases in membrane tension to their activation. Nevertheless, our findings are consistent with previous work to date that report both membrane tension decreases due to cell compression (7), as well as such compressive mechanostimulation resulting in mechanosensitive ion channel activation (38,44,72,73,74). More generally, any disturbance to the equilibrium distance between neighboring phospholipid molecules in the lipid bilayer gives rise to changes in its membrane tension (75). Similarly, asymmetric redistribution of lipids in the membrane, leads to alterations in its local curvature (76) and hence the clustering of the mechanosensitive ion channels, as observed in Fig. 3. Sufficient changes to both membrane tension or curvature, irrespective of whether they increase or decrease, for example, due to positive or negative pressure imparted on the membrane (77,78), can lead to activation of the mechanosensitive ion channels as long as the energy associated with such clustering and hence distortion to the membrane (for example, due to its invagination as a consequence of the decrease in membrane tension) and ion channel exceeds the conformational free energy of the ion channel $k_{\mathrm{B}}T$, wherein $k_{\mathrm{B}}$ is the Boltzmann constant and T the temperature (79). As such, our observations therefore reveal a possible general membrane mechanotransduction mechanism by which mechanosensitive ion channel activation can occur, not just due to the SRBW mechanostimulation, but also potentially with other types of stimuli: lipid microdomain redistribution as a consequence of deformation-driven membrane tension changes that leads to clustering and hence activation of the mechanosensitive ion channels. In addition, it potentially sheds light on a possible reason MscL, Piezo1, and transient receptor potential channels—membrane-associated mechanosensitive ion channels similar to Piezo1—have been reported to be activated at 30 and 7 MHz (22,80,81), respectively, the latter being shown to confer sensitivity of neuronal cells to mechanostimulation at that frequency.

As illustrated in Fig. 4 a, we thus postulate as follows the chain of events triggered by the SRBW mechanostimulation that leads to activation of the Piezo channels, and subsequently ${\text{Ca}}^{2+}$ mobilization within the cell to result in the observed biological fates. Dynamic changes in membrane tension as a result of the deformation to the plasma membrane induced by the SRBW forcing leads to transient increases in membrane order and polarization (i.e., increasing $L_o/L_d$ ratio), as manifested by the coalescence of the lipid rafts toward the nuclear regions of lower contractility and away from the invaginating $L_d$ regions (due to their lack of cytoskeletal structures) as the cell contracts (Fig. 3 a)—a consequence of the insufficient area to accommodate the lipid rafts in regions of large deformation since they cannot be invaginated into the cell. This provides an environment for their clustering with membrane-associated proteins, such as glycophosphatidylinositol-anchored proteins and mechanosensitive ion channels such as Piezo1 (Fig. 3 e). Activation of these Piezo channels is then a direct consequence of this crowding effect—only hypothesized theoretically to date (82) —together with changes in the membrane curvature and tension directly due to the SRBW-driven membrane deformation, through a force-from-lipid model (83,84,85). Simultaneously, anchoring of cytoskeletal actins to the lipid rafts via the membrane skeleton facilitates concurrent remodeling of the actin cytoskeletal network during SRBW microdomain reorganization (Fig. 3 d), which also allows for activation of the Piezo channels through a force-from-filament model (83,84). Together, the concomitant modulation of Ca²⁺ into the cell (Fig. 3 c) then allows for triggering of various downstream signaling cascades, such as the ALIX-mediated ESCRT, Rho-ROCK, and cAMP-mediated Epac-Rap1 pathways to result in the various downstream fates observed (12,16,18,20).

Figure 4.

High-frequency mechanotransduction and downstream cell fate. (a) Schematic illustration of the fundamental cellular (left) and (membrane) level mechanisms that govern the SRBW mechanotransduction process to yield diverse downstream cell fates that are (b) dependent on the modulation of ${\text{Ca}}^{2+}$ into the cell with the SRBW power, which, in turn, is strongly correlated with the $L_o/L_d$ ratio, as represented by the GP value associated with the membrane lipid order probe C-Laurdan with which the hMSCs are stained. A slight shift in lipid order at input powers between 0.5 and 1.5 W can be seen to initiate cytoskeletal reorganization, which, together with a 1.5-fold enhancement in intracellular ${\text{Ca}}^{2+}$, triggers appreciable increase in the biogenesis and exocytosis of extracellular vesicles (indicated by the arrows) due to initiation of the ESCRT pathway. At higher powers, the increase in $L_o$ microdomains, and their subsequent coalescence, yields a 2-fold increase in intracellular ${\text{Ca}}^{2+}$ that directs the hMSCs along an osteogenic differentiation pathway. Scale bars, 20 $\mu$m (exocytosis image) and 50 $\mu$m (cytoskeletal rearrangement and osteogenesis images).

Manipulation of lipid microdomains enables control of membrane-protein interactions and downstream cell fate

Fig. 4b shows that the downstream cell fate as a consequence of the SRBW mechanostimulation is connected to its ability to manipulate and hence spatiotemporally control lipid microdomain organization. More specifically, the different cell fates, namely exocytosis (exosome biogenesis and secretion), actin cytoskeletal remodeling, and osteogenic differentiation can be seen to be directed simply by varying the input SRBW power, given its ability to modulate the $L_o/L_d$ ratio (as represented by the GP values) and hence the intracellular Ca²⁺ concentration—a consequence of the activation of Piezo1, as discussed in the preceding section—that orchestrates the triggering of various different signaling cascades, such as the ESCRT, Rho-ROCK and Epac-Rap1 pathways. Moreover, it is also instructive to note that, even though the effect of the SRBW mechanostimulation on the membrane and ${\text{Ca}}^{2+}$ dynamics is transient, thereby resulting in ephemeral downstream effects (16), it can nevertheless also generate latent downstream effects through secondary signaling cascades involving cAMP (20) that are responsible for longer-term and even permanent effects (18).

To further highlight the potential implications of such direct spatiotemporal control over the lipid microdomains, we also demonstrate here the transformation of tau—a microtubule-binding intrinsically disordered protein, whose significance is pronounced under pathological conditions associated with neurodegenerative diseases—under high-frequency SRBW mechanostimulation. Fig. 5, in particular, shows the concentration of tau proteins along with lipid raft fusion (as evident from the higher $L_o/L_d$ ratio in Fig. 5 a–c, similar to that observed with the hMSCs in Fig. 2) in SH-SY5Y neuroblastoma cells near the cell nucleus when exposed to SRBW mechanostimulation (Fig. 5 d); these changes being dependent on the membrane composition, cholesterol content, and cytoskeletal structure induced by the SRBW forcing, all of which influence the modulation of ${\text{Ca}}^{2+}$ levels in the cell. Consistent with our previous observations, we also note the reversibility of the process, given the tendency for the tau protein to reorganize and the membrane to revert to its ground state after 15–30 min following cessation of the mechanostimulation. These results are further confirmed by the inhibitor studies in Fig. 5 e, where suppression of the SRBW-induced changes in the tau dynamics was observed in the presence of methyl-β-cyclodextrin to deplete the cholesterol content in the cell, and the membrane-permeable intracellular ${\text{Ca}}^{2+}$ chelator bis-acrylamide, BAPTA-AM, which acts to deplete the intracellular calcium store.

Figure 5.

SRBW-induced membrane lipid and membrane-associated protein distribution. (a) Images (inset) and distribution of their GP values ($-1$ to $+1$; as represented by the color bar) for untreated (control) and SRBW-irradiated (2.5 W for 5 min) C-Laurdan-labeled SH-SY5Y neuroblastoma cells fixed at 0 min postexposure incubation. The histograms for the GP values were fitted with Gaussian distributions, with the solid green curves representing the average between low (dashed blue curves) and high (dashed red curves) distributions. Scale bars, 30 $\mu$m. (b) High and (c) low GP coverage of both untreated and SRBW-irradiated cells, obtained by resolving the GP histograms in both cases with the two best Gaussian fits. (d) Lipid raft and tau protein organization in response to SRBW excitation (2.5 W for 5 min) at different postexposure incubation times, and (e) in the presence of BAPTA-AM and methyl-β-cyclodextrin (MβCD); the control comprised cells unexposed to SRBW stimulation but incubated with medium containing the same concentration in DMSO. Images were acquired at 100$\times$ and 60$\times$ magnification, respectively. Scale bars, 10 $\mu$m. Cell nuclei were stained using NucBlue Live ReadyProbes (blue), tau proteins with anti-tau antibody and Alexa Fluor 488 secondary antibody (green), and GM1 with Alexa Fluor 647 conjugate (red). The column on the right comprises a merger of these two channels. The data are represented in terms of mean $\pm$ standard error over triplicate runs, and the asterisks indicate statistically significant differences with $\ast p<0.05$, $\ast \ast p<0.01$ and $\ast \ast \ast \ast p<0.0001$, respectively.

We note that such accumulation of tau in the $L_o$ domains (the reduction in tau expression being attributed to SRBW-induced exocytosis), and their binding to GM1 and phosphatidylserine in the lipid rafts (Fig. 5 d), along with accompanying changes in cholesterol, intracellular ${\text{Ca}}^{2+}$, and the corresponding cytoskeletal remodeling, has been pathologically observed during many cases of neurodegeneration (86,87,88,89), where it has been shown that these membrane-tau interactions are accompanied by increased toxicity (32,33,90). In any case, this demonstration of membrane-tau interactions under the high-frequency mechanostimulation then attests to the amenability of the SRBW platform as a facile method for studying the membrane association dynamics of intrinsically disordered proteins, and, in doing so, opening up the possibility for understanding the complex cellular processes that drive tau dynamics under normal and pathological conditions, which, thus far, has been limited given that such studies have only been able to be conducted until now on model lipid bilayers (i.e., vesicles) (90,91,92,93).

Conclusions

By correlating changes to the membrane tension, and consequently, the plasma membrane lipid structure, to deformations to the cell imposed by high-frequency (10 MHz) mechanostimulation in the form of SRBWs, we have not only been able to explicate the fundamental mechanisms that explain the peculiar and unexpected response of cells to high-frequency MHz-order mechanostimulation but to also show, in addition, the possibility for direct, controlled manipulation of lipid microdomains—without the need for biochemical factors.

In particular, we show that decreases in the deformation-induced membrane tension as a consequence of the SRBW excitation lead to remodeling of the membrane lipid microdomain due to alterations in the membrane fluidity: with increasing SRBW power, we observe an increase in microdomain ordering (i.e., increase in $L_o/L_d$) and packing, which is accompanied by coalescence of the ordered $L_o$ microdomains and hence a decrease in their size. This has several consequences: 1) due to their stiffer gel-like state compared with the fluid-like $L_d$ microdomains, there is a concomitant shrinkage in the cell dimensions; 2) given the interconnectivity of the membrane skeleton with the cytoskeleton, the underlying actin cytoskeletal network is also altered; 3) clustering of lipid rafts rich in transmembrane proteins, leading to increased proximity and colocalization of ordered membrane-associated proteins (such as the mechanosensitive ion channels) with the $L_o$ microdomains, results in a crowding effect that is responsible for Piezo channel activation [It could perhaps be possible that a membrane mechanotransduction mechanism similar to that articulated here might be responsible for the activation of mechanosensitive ion channels in other studies, which had ruled out cavitation and thermal effects. For example, while Kubanek et al. (21) attributed alteration of the conformation states and hence activation of MEC-4-dependent ion channels to stretching of the plasma membrane as a consequence of the acoustic radiation pressure imparted on neuronal cells stimulated under 10 MHz ultrasound pulses, they did not explicitly describe the mechanotransduction mechanism that links the membrane stretching to ion channel activation. Moreover, Duque et al. (22), in fact, implicated the involvement of cholesterol and actin cytoskeletal interactions in the activation of transient receptor potential channels in human embryonic kidney 293T cells subjected to 1, 2, and 7 MHz ultrasound, which is consistent with that observed with the cholesterol-rich $L_o$ microdomain and cytoskeletal rearrangement in this work] (which has only been predicted theoretically to date (82)), that, in turn, mobilizes influx of the second messenger ${\text{Ca}}^{2+}$ into the cell to trigger a slew of signaling cascades (e.g., ESCRT, Rho-ROCK, and Epac1-Rap1) to direct the various downstream cell fates observed. These observations, together with the direct correlation between the $L_o/L_d$ ratio with the SRBW input power, therefore explicitly implicate the role of microdomain organization (and, more broadly, membrane order and polarization) on the determination of downstream cell fate.

In addition to the ability of the SRBW-driven spatiotemporal microdomain manipulation to influence intrinsically ordered membrane-associated proteins (e.g., Piezo1), we also show its potential for transformation of intrinsically disordered membrane-associated proteins, namely, tau proteins. Given that the aggregation of tau proteins and their association with the lipid membrane could be implicated in various neurodegenerative diseases, this first demonstration of such a unique possibility highlights the potential of the SRBW mechanostimulation platform as a facile tool to study membrane-tau interactions and the complex signaling cascades resulting from their association. We therefore envisage the platform to facilitate more thorough investigations into understanding the complex signaling processes accompanying the association of these proteins to the plasma membrane during pathological transformation that occurs during the progression of neurodegenerative diseases—an endeavor that has currently been limited due to the lack of a means for directly manipulating membrane microdomain organization in systems such as cells that are considerably more complex than the model bilayer systems that such manipulation has typically been demonstrated on to date.

Acknowledgments

L.Y.Y. and C.E.C. acknowledge support from the Australian Research Council through Discovery Project grant DP210101720.

Author contributions

Conceptualization, L.A.A. and L.Y.Y.; methodology, L.A.A., B.d.R., C.E.C., and L.Y.Y.; investigation, L.A.A.; visualization, L.A.A. and B.d.R.; supervision, L.Y.Y.; writing – original draft, L.A.A. and L.Y.Y.; writing – review & editing, L.A.A., B.d.R., C.E.C., and L.Y.Y.

Declaration of interests

Authors declare that they have no competing interests.

Editor: Guy Genin.